1 SAS Foundation Interfaces for Hadoop

Tool	Purpose	Product
FILENAME statement	Allow the DATA step to read and write HDFS data files.	Base SAS
PROC HADOOP	Copy or move files between SAS and Hadoop. Execute MapReduce and Pig code. Execute Hadoop file system commands to manage files and directories.	Base SAS
SQL Pass-Through	Submit HiveQL queries and other HiveQL statements from SAS directly to Hive for Hive processing. Query results are returned to SAS.	SAS/ACCESS Interface to Hadoop
LIBNAME Statement For Hadoop	Access Hive tables as SAS data sets using the SAS programming language. SAS/ACCESS engine translates SAS language into HiveQL and attempts to convert the processing into HiveQL before returning results to SAS.	SAS/ACCESS Interface to Hadoop

2 SAS In-Memory Analytics Interfaces for Hadoop

Hadoop is now one of the file storage systems that SAS uses for SAS In-Memory Analytics product solutions:
- SAS High-Performance Analytics products
- SAS Visual Analytics
- SAS In-Memory Statistics
- SAS Code Accelerator for Hadoop (DS2)

2.1 A Hadoop Cluster to Run SAS

2.2 Base SAS: FILENAME for Hadoop

2.3 Base SAS: The Same Process for PROC HADOOP

2.4 SAS/ACCESS: SQL Pass-Through and LIBNAME

2.5 In-Memory Analytics

2.6 Computing Environment

3 Managing Files and Executing Hadoop Commands, Map-Reduce, and Pig

Part 1: Introduction to Base SAS Methods for Hadoop
Part 2: The HADOOP FILENAME Statement and PROC HADOOP
Part 3: Executing Pig Code With PROC HADOOP

4 Part 1: Introduction to Base SAS Methods for Hadoop

4.1 Base SAS Tools for Hadoop

The Hadoop FILENAME statement enables you to do the following:
- upload local data to Hadoop using the DATA step
- read data from Hadoop using the DATA step
PROC HADOOP enables you to do the following:
- submit Hadoop file system (HDFS) commands
- submit MapReduce programs
- submit Pig script

4.2 The Hadoop `Config.xml` File

The FILENAME statement for Hadoop and PROC HADOOP require an option that specifies a Hadoop configuration file (config.xml).
- The configuration file defines how to connect to Hadoop.
- The file must be accessible to the SAS client application.
- A SAS administrator commonly manages this configuration for the SAS users.
- This file is often referred to as the Hadoop core-site.xml file.

4.3 Hadoop JAR Files

A collection of Hadoop JAR files is also required on the SAS client machine.
- An environment variable SAS_HADOOP_JAR_PATH on the SAS client machine defines the location of the Hadoop JAR files
- The Hadoop JAR files must be compatible with the specific Hadoop implementation and can be copied from the Hadoop server.
- A Hadoop system administrator commonly manages the configuration of the Hadoop JAR files for the SAS users.

4.4 Base SAS Interface to Hadoop

5 Part 2: The HADOOP FILENAME Statement and PROC HADOOP

5.1 Business Scenario

We want to develop a prototype for a process that uses SAS to orchestrate the following scenario:
1. Move unstructured text files into the Hadoop file system.
2. Invoke MapReduce programs developed by Java programmers in order to:
  - read and process the text files to perform various analyses
  - output results as text files in the Hadoop file system
3. Read the summarized text analysis results back into SAS for further analysis and reporting purposes.

5.1.1 Breakdown 1

Move unstructured text files into the Hadoop file system.

PROC HADOOP and the HDFS COPYFROMLOCAL statement

5.1.2 Breakdown 2

Invoke MapReduce programs developed by Java programmers in order to:
- read and process the text files to perform various analyses
- output results as text files in the Hadoop file system

PROC HADOOP and the MAPREDUCE statement

5.1.3 Breakdown 3

Read the summarized text analysis results back into SAS for further analysis and reporting purposes.

FILENAME statement for Hadoop and DATA step

5.2 Business Scenario Pseudocode

proc hadoop...;
  hdfs copyfromlocal="local file" out="hdfs file";
run;

proc hadoop...;
  mapreduce input="hdfs file" output="hdfs outfile" ...;
run;

filename fileref "hdfs outfile"...;
  data somedata;
  infile fileref input...;
  ...
run;

5.2.1 Breakdown 1

Move unstructured text files into the Hadoop file system

proc hadoop...;
  hdfs copyfromlocal="local file" out="hdfs file";
run;

5.2.2 Breakdown 2

Input the hdfs file to MapReduce program and output results to hdfs file

proc hadoop...;
  mapreduce input="hdfs file" output="hdfs outfile" ...;
run;

5.2.3 Breakdown 3

Read the MapReduce output with SAS for further processing

filename fileref "hdfs outfile"...;
  data somedata;
  infile fileref input...;
  ...
run;

5.3 PROC HADOOP

PROC HADOOP submits:
- Hadoop file system (HDFS) commands
- MapReduce programs
- PIG language code

PROC HADOOP <Hadoop-server-option(s)>;
  HDFS <Hadoop-server-option(s)> <hdfs-command-option(s)>;
  MAPREDUCE <Hadoop-server-option(s)> <mapreduce-option(s)>;
  PIG <Hadoop-server-option(s)> <pig-code-option(s)>;
  PROPERTIES <configuration-properties>;
RUN;

5.4 HDFS Statements

HDFS COPYFROMLOCAL="local-file" OUT="output-location" <DELETESOURCE> <OVERWRITE>;

HDFS COPYTOLOCAL="HDFS-file" OUT="output-location" <DELETESOURCE> <OVERWRITE> <KEEPCRC>;

HDFS DELETE="HDFS-file" <NOWARN>;

HDFS MKDIR="HDFS-path";

HDFS RENAME="HDFS-file" OUT="new-name";

5.5 Moving a File from SAS to Hadoop

This program creates a directory in the Hadoop file system (HDFS) and copies a file from the SAS server to the new HDFS directory.

filename hadconfg "/work/hadoop_config.xml";
proc hadoop options=hadconfg username="&std" verbose;
  hdfs mkdir="/user/&std/data";
  hdfs copyfromlocal="/work/DIACCHAD/data/moby_dick_via_sas.txt" out="/user/&std/data";
run;

5.6 Execute MapReduce Code

MAPREDUCE <Hadoop-server-option(s)> <mapreduce-option(s)>;

proc hadoop options=hadconfg username="&std";
  mapreduce
    input="source-file"
    output="target-file"
    jar="jar file containing MapReduce code"
    outputkey="output key class (in MapReduce code)"
    outputvalue="output value class"
    reduce="reducer class"
    combine="combiner class"
    map="map class";
run;

5.6.1 Breakdown 1

input="source-file"

The input file in HDFS the MapReduce program reads

output="target-file"

The output file in HDFS the MapReduce program writes to

5.6.2 Breakdown 2

jar="jar file containing MapReduce code"

The JAR file containing the MapReduce program and named classes

5.6.3 Breakdown 3

outputkey="output key class (in MapReduce code)"

The name of the output key class in dot notation

outputvalue="output value class"

The name of the output value class in dot notation

5.6.4 Breakdown 4

reduce="reducer class"
combine="combiner class"
map="map class";

The Java classes in the map reduce program that execute the map, reduce and combine steps

5.7 MapReduce Example

In the demonstration, PROC HADOOP will be used to invoke a MapReduce program that will do the following:
- read a text file containing free unstructured text. This file can be distributed in multiple data nodes in Hadoop
- parse the text into the individual words in each data node
- count up the number of instances of each unique word found in each data node
- combine total counts for each unique word across nodes
- write the results to an HDFS output file

proc hadoop options=hadconfg username="&std" verbose;
  mapreduce
    jar = "<hdfs path>/hadoop-mr1-cdh.jar"
    input = "<hdfs path>/moby_dick_via_sas.txt"
    map = "org.apache.hadoop.examples.WordCount$TokenizerMapper"
    reduce = "org.apache.hadoop.examples.WordCount$IntSumReducer"
    combine = "org.apache.hadoop.examples.WordCount$IntSumReducer"
    outputkey = "org.apache.hadoop.io.Text"
    outputvalue = "org.apache.hadoop.io.IntWritable"
    output = "<hdfs path>/mapoutput"
  ;
run;

5.7.1 Breakdown 1

Read a text file containing free unstructured text

input = "<hdfs path>/moby_dick_via_sas.txt"

5.7.2 Breakdown 2

In parallel in each data node, parse the text into the individual words

map = "org.apache.hadoop.examples.WordCount$TokenizerMapper"

5.7.3 Breakdown 3

In each data node, in parallel, count up the number of instances of each unique word found

reduce = "org.apache.hadoop.examples.WordCount$IntSumReducer"

5.7.4 Breakdown 4

Combine the counts for each unique word across data nodes to find final counts

combine = "org.apache.hadoop.examples.WordCount$IntSumReducer"

5.7.5 Breakdown 5

The output contains each unique word (outputkey) as Text and the number of times it occurred in the input file as Integer

outputkey = "org.apache.hadoop.io.Text"
    outputvalue = "org.apache.hadoop.io.IntWritable"

5.7.6 Breakdown 6

The HDFS location the output is written to

output = "<hdfs path>/mapoutput"

5.8 The FILENAME Statement for Hadoop

In SAS, the FILENAME statement associates a fileref with an external file and an output device or access method.

FILENAME fileref <device type or access method> "external file" <options>;

filename in hadoop "Hadoop-file-path" concat cfg=xml-config-file user="&std";

5.8.1 Breakdown 1

in

file reference

5.8.2 Breakdown 2

hadoop

Access method

5.8.3 Breakdown 3

"Hadoop-file-path"

The directory containing the concatenated files to read

5.8.4 Breakdown 4

concat

Specifies to read each file in the directory defined by the Hadoop file path
Note: This option is valid only for reading (not writing) Hadoop files with the FILENAME statement.

5.8.5 Breakdown 5

cfg=xml-config-file

Points to the location of the Hadoop configuration file on the machine where SAS is executing.

5.8.6 Breakdown 6

user="&std"

The user ID to connect to Hadoop

5.9 Reading a Hadoop File with a DATA Step

filename hadconfg "/work/hadoop_config.xml";

filename mapres hadoop "/user/&std/data/mapoutput" concat cfg=hadconfg user="&std";

data work.commonwords;
  infile mapres dlm="09"x;
  input word $ count;
  ...
run;

hadconfg <=====> cfg=hadconfg
mapres <=====> infile mapres
"09"x is the hex code constant for the tab character

5.9.1 Another Example

filename hadconfg "/work/hadoop_config.xml";
filename orders hadoop "/user/shared/data/custorders.txt" cfg=hadconfg user="&std";

data work.custorders;
  infile orders;
  input @1 customer_id 8. ...;
run;
proc print data=work.custorders;
run;

Does the DATA step read a single HDFS file or a concatenated directory?

5.10 Base SAS: FILENAME for Hadoop (Review)

5.11 Reading a Hadoop File with a DATA

5.11.1 Step 1

5.11.2 Step 2

5.11.3 Step 3

5.12 Writing a Hadoop File with a DATA

filename hadconfg "/work/hadoop_config.xml";
filename out hadoop "/user/&std/data/custord" dir cfg=hadconfg user="&std";

data _null_;
  set work.custorders;
  file out(corders) dlm=",";
  put customer_id
      country
      gender
      birth_date
      product_id
      order_date
      quantity
      costprice_per_unit;
run;

dir: To write to files in the directory specified by the Hadoop file path
corders: Create a file called corders in the HDFS directory "/user/&std/data/custord"

5.12.1 Step 1

5.12.2 Step 2

5.12.3 Step 3

5.13 Use Ambari to Browse the Hadoop File System

6 Part 3: Executing Pig Code With PROC HADOOP

6.1 Executing Pig Code with PROC HADOOP

PIG CODE=fileref | "external-file"
    PARAMETERS=fileref | "external-file"
    REGISTERJAR="external-file(s)"
;

filename pigcode "/workshop/DIACCHAD/pigcode.txt";

proc hadoop options=hadconfg username="hdfs" verbose;
  pig code=pigcode;
run;

pigcode.txt:

A = LOAD '/user/shared/data/custord'
    USING PigStorage (',')
    AS (customer_id, country, gender, birth_date, product_id,
        order_date, quantity, costprice_per_unit);

B = FILTER A BY gender == 'F';

store B into '/user/shared/data/student1';

6.1.1 Pig Code Breakdown 1

A = LOAD '/user/shared/data/custord'
    USING PigStorage (',')
    AS (customer_id, country, gender, birth_date, product_id,
        order_date, quantity, costprice_per_unit);

Load the comma delimited file and name each field

6.1.2 Pig Code Breakdown 2

B = FILTER A BY gender == 'F';

Subset the loaded file for records where gender = ‘F’

6.1.3 Pig Code Breakdown 3

store B into '/user/shared/data/student1';

Store the results in the Hadoop file system directory indicated

7 Using the SQL Pass-Through Facility

7.1 SAS/ACCESS Interface to Hadoop

7.2 SQL Pass-Through Query Example

proc sql;
  connect to hadoop (server=namenode port=10000 subprotocol=hive2 schema=diacchad user="&std");
    select * from connection to hadoop
      (select employee_name, salary
        from salesstaff
        where emp_hire_date between '2011-01-01' and '2011-12-31'
      );
  disconnect from hadoop;
quit;

Note: The query in the bracket is sent directly to Hive and is executed as a HiveQL query by Hive.

7.3 Joining Multiple Tables from Different Databases

7.4 Creating a SAS File from Hive Results

7.5 Using HiveQL DDL Statements in SAS

8 Using the SAS/ACCESS LIBNAME Engine

Using the LIBNAME Statement for Hadoop
Using Data Set Options
Creating Views
Combining Tables
Transferring Data Sets from SAS to Hive

9 The LIBNAME Statement (Review)

The LIBNAME statement assigns a libref to a SAS library.

LIBNAME libref 'SAS-data-library' <options>;

Rules for naming a libref:
- The name must be eight characters or less.
- The name must begin with a letter or underscore.
- The remaining characters must be letters, numbers, or underscores.

10 The SAS/ACCESS LIBNAME Statement

The SAS/ACCESS LIBNAME statement does the following:
- establishes a libref, which acts as an alias or nickname to Hive.
- permits a Hive table to be referenced by a two-level name.
- enables the use of the SAS/ACCESS LIBNAME statement options to specify how Hive objects are processed by SAS.
- enables you to customize how to connect to Hive.

libname hivedb hadoop server=namenode
        subprotocol=hive2
        port=10000 schema=diacchad
        user=studentX pw=StudentX;

LIBNAME libref engine-name <connection-options>
<LIBNAME-options>;

SAS Log

23    libname hivedb hadoop server=namenode
24        subprotocol=hive2
25        port=10000 schema=diacchad
26        user="&std" pw="&stdpw";
NOTE: Libref HIVEDB was successfully assigned as follows:
  Engine: HADOOP
  Physical Name: jdbc:hive2://namenode:10000/diacchad

STSCI 5065 - Lecture 20

Hadoop and SAS Integration

Harry Lu

1 SAS Foundation Interfaces for Hadoop

2 SAS In-Memory Analytics Interfaces for Hadoop

2.1 A Hadoop Cluster to Run SAS

2.2 Base SAS: FILENAME for Hadoop

2.3 Base SAS: The Same Process for PROC HADOOP

2.4 SAS/ACCESS: SQL Pass-Through and LIBNAME

2.5 In-Memory Analytics

2.6 Computing Environment

3 Managing Files and Executing Hadoop Commands, Map-Reduce, and Pig

4 Part 1: Introduction to Base SAS Methods for Hadoop

4.1 Base SAS Tools for Hadoop

4.2 The Hadoop Config.xml File

4.3 Hadoop JAR Files

4.4 Base SAS Interface to Hadoop

5 Part 2: The HADOOP FILENAME Statement and PROC HADOOP

5.1 Business Scenario

5.1.1 Breakdown 1

5.1.2 Breakdown 2

5.1.3 Breakdown 3

5.2 Business Scenario Pseudocode

5.2.1 Breakdown 1

5.2.2 Breakdown 2

5.2.3 Breakdown 3

5.3 PROC HADOOP

5.4 HDFS Statements

5.5 Moving a File from SAS to Hadoop

5.6 Execute MapReduce Code

5.6.1 Breakdown 1

5.6.2 Breakdown 2

5.6.3 Breakdown 3

5.6.4 Breakdown 4

5.7 MapReduce Example

5.7.1 Breakdown 1

5.7.2 Breakdown 2

5.7.3 Breakdown 3

5.7.4 Breakdown 4

5.7.5 Breakdown 5

5.7.6 Breakdown 6

5.8 The FILENAME Statement for Hadoop

5.8.1 Breakdown 1

5.8.2 Breakdown 2

5.8.3 Breakdown 3

5.8.4 Breakdown 4

5.8.5 Breakdown 5

5.8.6 Breakdown 6

5.9 Reading a Hadoop File with a DATA Step

5.9.1 Another Example

5.10 Base SAS: FILENAME for Hadoop (Review)

5.11 Reading a Hadoop File with a DATA

5.11.1 Step 1

5.11.2 Step 2

5.11.3 Step 3

5.12 Writing a Hadoop File with a DATA

5.12.1 Step 1

5.12.2 Step 2

5.12.3 Step 3

5.13 Use Ambari to Browse the Hadoop File System

6 Part 3: Executing Pig Code With PROC HADOOP

6.1 Executing Pig Code with PROC HADOOP

6.1.1 Pig Code Breakdown 1

6.1.2 Pig Code Breakdown 2

6.1.3 Pig Code Breakdown 3

7 Using the SQL Pass-Through Facility

7.1 SAS/ACCESS Interface to Hadoop

7.2 SQL Pass-Through Query Example

7.3 Joining Multiple Tables from Different Databases

7.4 Creating a SAS File from Hive Results

7.5 Using HiveQL DDL Statements in SAS

8 Using the SAS/ACCESS LIBNAME Engine

9 The LIBNAME Statement (Review)

10 The SAS/ACCESS LIBNAME Statement

11 LIBNAME Statement Connection Options

12 Listing of Hive Tables in the Schema

13 Contents of a HiveTable

14 Listing of a Hive Table

15 SAS/ACCESS Engine Implicit Pass-through

16 Implicit versus Explicit SQL Pass-through

4.2 The Hadoop `Config.xml` File